Personal injury firms manage reviews as a number: get the stars up, get the count up, put the badge on the homepage. Meanwhile, the systems deciding visibility stopped reading reviews as numbers years ago. Google holds a family of patents on sentiment detection as a ranking signal for reviewable entities, US 8417713 and its continuations through US 10394830: machinery that reads what reviewers WRITE, extracts the sentiment and the substance, and uses it to rank the entities being reviewed. The words are the signal. The stars are the summary the machines no longer need. This piece is about operating reviews accordingly: as a written, machine-read, legally bounded public record of what your firm is actually like.
Stars stopped discriminating. In my audit of 1,005 page-one PI firms, the median firm already held dozens of reviews, and the vertical clusters so tightly near the top of the scale that ratings barely separate anyone: nearly everyone competing seriously looks like a 4.6 to 5.0. When every firm on the shortlist wears five stars, the stars cannot be the decision, for the machine or the human.
The text still discriminates enormously. Two five-star firms read nothing alike: one's reviews mention the named attorney, the case type, the communication, the outcome experience; the other's say "great service" forty times, in suspiciously similar cadence. The sentiment-detection patents describe extracting exactly this: polarity and substance from review language, attached to the entity, usable as signal. And the human shortlist-maker performs the same extraction manually at 11pm, reading your newest ten reviews to find out what you are like when nobody is marketing.
Both audiences read the words, one at scale and one at midnight, and both decide accordingly. Operate the words.
You cannot write your clients' reviews, and in this profession you should not try to steer them with scripts; between bar advertising rules and the FTC's rule against fake and incentivized reviews, the compliance lane is narrow and the penalties are real. What a firm CAN do, honestly, is shape the conditions under which true words get written.
Ask at the moment the substance is fresh. The natural moment is case resolution, when the client can speak to the whole arc. The ask is a simple, personal request from the human who handled them, not a blast from a tool, and never conditioned on sentiment.
Make the ask open-ended toward specifics. Not "leave us five stars," which is coaching the summary, but "if you are willing, share what the experience was actually like." Clients who write specifics write the case type, the attorney's name, the communication, the fears that got handled, which is precisely the substance the extraction machinery and the midnight reader both feed on.
Let the mix be real. A wall of identical praise reads as manufactured to both audiences; one patent-family insight the industry ignores is that systems built to read sentiment are necessarily built to notice distributional weirdness. A firm with overwhelmingly strong, varied, specific reviews and the occasional imperfect one, answered well, reads as a real place. Which it is.
Every reply a firm posts is public, permanent, and read by three audiences at once: the reviewer, every future prospect, and the extraction machinery parsing the thread. Most firms either reply never, reply with a template, or reply only to attack their one-star reviews, which is the worst possible distribution: silence toward friends, heat toward critics, in public, forever.
The operating standard is professional and boring on purpose. Thank the specific, by name where appropriate. Answer the unhappy with calm process and an offline path, never case details; confidentiality does not pause for a bad review, and a defensive reply breaching client information is a bar complaint that started as marketing. Write every reply as if it will be read aloud someday by someone deciding whether to trust you, because that is literally its function.
A quiet operational truth from audit work: reply behavior is one of the fastest reads on whether anyone owns a firm's public record at all. Unanswered reviews, like unanswered forms, are usually the visible corner of a wholly unowned layer.
One more property of the record that star-count thinking misses entirely: reviews age, and the reader's attention does not reach the archive. The default sort puts the newest at the top, the shortlist-maker reads the first screen or two, and that means your operating reputation is effectively your most recent handful of reviews, whatever the lifetime count says. A firm with 300 reviews whose newest one is fourteen months old is presenting a museum; the reader's quiet inference is that either the firm stopped winning clients or stopped caring what they say, and neither inference helps at midnight.
This is why the honest ask, built into the case-resolution workflow rather than run as an occasional campaign, beats every burst strategy. A steady few reviews a month, every month, keeps the visible layer current and reads as a living practice. The burst pattern, forty reviews in March after a vendor's "reputation sprint" and silence since, is legible as a campaign to any human who scrolls the dates, and dates are one of the easiest features for the extraction machinery to read. Steadiness is not just safer; it is the only pattern that looks like what it claims to be, because it is.
The recency lens also tells you where to spend attention when resources are thin: you do not need to answer a decade of backlog this week. Start with everything from the last ninety days, keep current from today forward, and work backward as time allows. The newest screen is the storefront; sweep it first.